Empirical Research: Emerging Research: Robust and Efficient Learning: Modeling and Remediating Students' Domain Knowledge

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

This project employs an intelligent learning environment called the Genetics Cognitive Tutor to explore student learning via abductive problem solving in undergraduate genetics courses. Problem solving is an essential learning activity across STEM domains, but there is a well-documented risk that students can develop superficial knowledge based on surface features of the problem situation, instead of a deeper conceptual understanding. The key hypotheses of this project are that shallow reasoning emerges because students lack the conceptual grounding to engage in deep reasoning, and that this shallow reasoning can be detected during problem solving and remediated as soon as it occurs. This empirical research project has two chief goals to promote deeper reasoning in genetics problem solving. First, the project is developing new on-line Conceptually Grounded Learning Activities to better prepare students for deeper learning in subsequent problem solving. Second, the project employs cognitive modeling and machine learning techniques to develop a model for tracing student knowledge that distinguishes between superficial and deep reasoning in real time during problem solving. This permits the Cognitive Tutor to intervene appropriately as superficial reasoning occurs, by engaging students in reasoning directly about underlying domain concepts. The new learning activities and knowledge tracing technique are being evaluated and refined in laboratory studies in universities where genetics is taught, and expanded to authentic classroom-based settings at four diverse universities that vary by institutional types across four dimensions: public vs. private; national vs. regional; student academic proficiency; and ethnic diversity. The dependent measures of deep vs. superficial reasoning include response time, accuracy and response history in the tutor, and post-test measures of retention, transfer, and preparation for future learning. This research targets a critical need in undergraduate biology education. Genetics is a linchpin of biology instruction, both because it is a fundamental, unifying theme of biology, and because it is viewed by students and instructors as one of the most challenging topics in biology. The research will directly inform improvements in intelligent learning environments, such as the Genetics Cognitive Tutor, to promote deeper reasoning in genetics problem solving. But shallow reasoning is a challenge across STEM domains and across other types of learning activities. The knowledge developed in this project should lead directly to design guidelines for intelligent problem-solving environments that support conceptually-grounded reasoning in other STEM domains. The lessons learned can also guide researchers in creating shallow-reasoning detectors in a broad range of on-line learning environments, and can provide guidelines for sequencing different types of learning activities in non-computerized learning environments.